The LLM Critics Are Right. I Use LLMs Anyway
Points and comments are a snapshot, not live.
A software engineer admits LLM critics are right but still spends $10k/month on tokens.
The author agrees with all major LLM criticisms: slop, copyright, environmental harm, bubble risk, geopolitical fragility, and the erosion of trust in open-source contributions and junior mentoring. Yet he spends $10,000/month on tokens to amplify his own thinking. He argues that human credibility behind AI use is the key differentiator, and describes techniques like the "grill me" skill and Basecamp-style problem statements to maintain quality. At Local-First Conf, he observed even speakers who criticized LLMs had Claude Code open, creating a widespread dissonance.
What commenters are saying
Commenters split into two camps: those who see inevitable skill atrophy from outsourcing coding to LLMs, and those who find LLMs accelerate learning in unfamiliar areas. A recurring concern is that you cannot ethically delegate tasks you cannot do yourself, citing regex as an example. One commenter notes top-down learners benefit more from LLMs while bottom-up learners find them suspicious. Another suggests deliberately choosing not to outsource some tasks to maintain skills.